Two robust observations by psychopathologists interested in the origins and maintenance of anxiety-related disorders led to the present proposal. First, it has been repeatedly demonstrated that Pavlovian aversive conditioning preparations reliably change the fear-eliciting threat value of a wide range of stimuli, and that such changes influence, and are influenced by, automatic (i.e., involuntary, unaware) and strategic (i.e., deliberate, effortful) cognitive processing. Second, using several methods derived from cognitive science, researchers have consistently shown that persons suffering from anxiety-related disorders and non-clinical anxious conditions show both automatic and strategic preferential biases for processing threatening or unpleasant stimuli that are specific to their condition. The most common application of cognitive paradigms is to demonstrate that information-processing bias exists in a disorder, and the majority of studies have been concerned with the generalization of such biases across different patient groups, or different stimulus materials. Yet, no studies have demonstrated that such biases do, in fact, arise from a history of negative emotional learning, despite the implicit assumption that such learning is, in part, responsible for threat-biased processing. The purpose of the four experiments proposed here is to establish and elucidate the relation between negative emotional learning and attentional and automatic threat-biased processing. Negative emotional learning will be established using an aversive Pavlovian differential conditioning preparation, with the conditional stimuli (CSs) being innocuous words which will be paired or unpaired with unconditioned stimuli (UCSs) consisting of panicogenic doses of carbon dioxide-enriched air. Autonomic and self-reported indices of conditioned and unconditioned responding will be assessed before, during, and following conditioning. Stroop color-naming and lexical/repetition priming tasks will be embedded within conditioning phases to assess differential acquisition of threat- biased processing, and the extent to which such biases can be readily attenuated following extinction. Experiment l will elucidate the relation between direct and verbally-mediated negative-emotional learning and attentional threat bias, whereas Experiment 2 will address whether such biases are-automatic. Experiments 3a and 3b will establish whether extent of threat- biased processing varies reliably with strength of negative- emotional learning (conditioning), and if such effects can be either potentiated or depotentiated with the passage of time between learning and performance. Unpaired and unseen control words will reveal whether the predicted biases are specific to threat or whether they extend to any emotional information. The findings ought to (a) clarify the relation between threat-biased processing and negative emotional learning, while (b) elucidating the conditions under which emotional responding may be decoupled from biases in cognitive processing.